-
YART 11.- HETEROGENEOUS REACTIONS.
CHAIRMAN.-PrOf. F. G. DONNAN, C.B.E., PH.D., F.R.S.
CHEMICAL REACTIONS ON SURFACES.
OPENING ADDRESS BY DR. IRVING LANGMUIR.
,After the discovery of the law of mass action, and its kinetic
interpreta- tion, it was at first taken for granted that the same
principle would apply unaltercd to heterogeneous reactions; that
is, it was assumed that the reaction velocity of a substance in
contact with a solid would be propor- tional to the concentration
of one or more of the reacting substances. Subsequent work showed
that other factors than the mere mass action effect were important
in determining the velocity of these reactions.
that the rate of solution of solid substances in liquids is
often limited by the rate of diffusion of the dissolved substances
away from the surface. At this surface, therefore, the solution
remains practically saturated at all times.
Nernst extended this theory to cover heterogeneous reactions in
general. H e assumed that all solid surfaces were covered with
adsorbed films, and that the reacting substaxces must diffuse
through these films before coming in contact with the underlying
nietal or other substance constituting the solid. H e assumed that
in general the rate of reaction was limited by this diffusion and
that the reaction would be practically instantaneous if it were not
for the zdsorbed film.
Eodenstein and Fink adopted the general features of this theory,
but considered that the film varied in thickness, depending upon
the partial pressure of the gases in contact with the solid. I n
this way they were able to account for cases where the reaction
velocity is not proportional to the concentrations of the reacting
substances. For example, it was found ex- perimentally that the
velocity of the oxidation of sulphur dioxide with a platinum
catalyser, as in the contact process, was inversely proportional to
the square root of the pressure of the sulphur trioxide. They
explained this by assuming that the platinum was covered by an
adsorbed film of SO,, whose thickness was proportional to the
square root of the pressure of this component.
,Although this theory suggests a possible mechanism for the
effect of catalytic poisons, it has not proved to be a satisfactory
general theory of catalytic action. Thus, there is no logical
reason for assuming, in some reactions, that the thickness of the
adsorbed film is proportional to the square root of the pressure,
while in other reactions, it is proportional to the first power of
the pressure.
These theories of diffusion through films require the existence
of films
It was shown by Noyes and Whitney
12. physik. Cheiiz., 23, 689 (1897). Ibid., 60, 46 (1907).
607
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online / Journal Homepage / Table of Contents for
this issue
-
Go8 CHEMICAL REACTIONS ON SURFACES
relatively thick in proportion to the dimensions of molecules,
for we find experimentally that the reaction velocities can vary a
thousand, if not a millionfold, in reactions where we have to
account for this change by a variation in the thickness of a film.
I n such cases it would be necessary to have films so thick that we
should be able to see them. Fink, however, measured the amount of
SO, adsorbed by the platinum per unit area, and found it to be of
the order of magnitude of a single layer of molecules. I t is,
then, hardly logical to assume that the thickness of this film can
vary in proportion with the square root of the pressure for a wide
range of pressures.
Euidence fo. the Exz'steizce of very StnbZe Adsorbed Fidms.
-Experiments which the writer began in 1912 showed that the effect
of residual gases on the electron emission from heated tungsten
filaments in vacuum was generally to decrease the emission, instead
of to increase it. Oxygen, or traces of water vapour, had a really
remarkable effect in decreasing the current. Thus, at temperatures
of about 1900~ E., the emission was de- creased many thousandfold
by pressures of oxygen as low as one bar (one dyne per square
centimetre, or approximately 10-6 atmospheres). I t did not seem
possible that the oxygen could prevent the emission of the
electrons unless it covered in some form the larger part of the
surface. This film, however, must have been an extraordinarily
stable one, to remain on a filament in such good vacuum a t this
high temperature. At tempera- tures even as low as 1000' K. no
visible film is formed on the surface of tungsten by introducing
oxygen, for the WO, which is produced distils off and leaves the
surface apparently clean.
Since that time a long series of investigations has been made on
the effect of low pressures of oxygen in altering the properties of
tungsten a t high temperatures. All of this work confirms the view
that even a t the highest temperatures, in the presence of traces
of oxygen, the surface of the filament is practically completely
covered with a film of oxygen.
Thus, when the filament is heated to 3 3 0 0 ~ K. and a pressure
of oxygen of a few bars is admitted to the bulb, the rate of
disappearance of the oxygen shows that about 50 per cent. of all
the oxygen molecules which strike the filament react with it to
form W03, which distils on to the bulb. Since there are three atoms
of oxygen in the molecule of this compound and only two in the
oxygen molecule, it is clear that at least one-half of the tungsten
surface, even at this high temperature, must be covered with oxygen
in some form.
The chemical effects of this adsorbed oxygen film are as
striking as the effects on the electron emission. If a tungsten
filament is heated to 1500~ K., or more, in pure, dry hydrogen at
low pressure, the hydrogen is gradually dissociated into atoms and
the atomic hydrogen is adsorbed by the glass walls of the vessel or
reacts with any WO, which may previously have been distilled on to
the bulb. The hydrogen pressure therefore gradually decreases. This
effect is entirely prevented by minute traces of oxygen. Thus, if a
mixture of oxygen and hydrogen be introduced into a bulb and the
filament heated to I ~ O O O , instead of the gases reacting to
form water vapour, as they would in contact with a platinum
filamect, the oxygen reacts gradually with the tungsten to form
W03. While this is going on, the dissociation of the hydrogen by
the filament is entirely prevented, so that finally nearly pure
hydrogen remains and the pressure becomes constant. After ten or
fifteen minutes the pressure of the oxygen decreases to such a
point (a minute fraction of one bar) that it no longer is able to
prevent the dissociation of the hydrogen. This then begins suddendy
to dissociate, and in a.few minutes more all of the hydrogen has
disappeared.
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL REACTIONS ON SURFACES 609
The oxygen film on the tungsten surface thus consists of oxygen
in a: form which cannot react with hydrogen even at 1500'. I t
certainly does not behave like a layer of either tungsten oxide or
of highly compressed oxygen gas. Its chemical properties have been
completely modified by its adsorption on the tungsten.
The function of the oxygen in preventing the dissociation of the
hydro- gen is clearly that of a cata/yticpoison. This effect of the
oxygen on tung- sten is observed with several other reactions. For
example, methane is decomposed by tungsten, giving hydrogen, while
the carbon is taken up by the tungsten filament, but if the methane
is mixed with oxygen, it is not decomposed until all the oxygen has
reacted with the tungsten to form WO, and it is then decomposed as
though no oxygen had been present. The same thing happens with
ammonia, which, alone, is decomposed easily by a tungsten filament
at goo0 K., but in presence of oxygen is not decomposed unless the
filament temperature is raised above about 1300' K.
If the electron emission is measured while a mixture of hydrogen
and oxygen is in contact with the filament, it is found that the
electron emission increases suddenly at the same instant that the
dissociation of the hydrogen begins.
The remarkable stability of these oxygen films, as well as the
complete change in the chemical properties of the oxygen, gives
reason for believing that the surface is covered with individual
oxygen atoms chemically com- bined with the underlying tungsten
atoms. This film cannot be regarded as consisting of an oxide of
tungsten, nor as atomic oxygen, in the sense in which we think of
free oxygen atoms. The oxygen Atoms are probably held to the
surface by four pairs of electrons, just as the oxygen atom is held
to the carbon atom in CO,. The oxygen atoms are thus chemically
saturated, but the tungsten atoms are not saturated, so that they
are held by strong forces to the tungsten atoms that lie below
them. This kind of structure is quite in accord with the conception
of the struc- ture of solids to which w e are led by the work of
the Braggs, on crystal structure.
Work with other metals has shown that stable films of the kind
we have just been discussing are of very common occurrence. Oxygen
forms a similar film on carbon, and carbon monoxide, hydrogen,
cyanogen, hydro- gen sulphide, phosphine, and arsine form stable
films on platinum. I t is probable that all substances that have a
poisoning effect on catalytic surfaces form films of this kind.
Erdidence that these Stable FiZm are MonomokcuZa'ar and that the
&foZe- cuZes tend to be Oriented on the Surface.-According to
our present concep- tions, atoms consist of electrons arranged in
space about a positively charged nucleus. Whether we assume that
the electrons are moving or not, it is certain that the electrons
nearly completely surround the nucleus. In most molecules, atoms
share pairs of electrons (duplets) with each other. I n any
electrically neutral molecule, the field of force must decrease in
in- tensity with a very high power of the distance from the centre.
Bohr has calculated that the electric force around a group of eight
electrons, arranged at the corners of a cube and surrounding a
nucleus having an equivalent charge, is inversely proportional to
the tenth power of the distance from the nucleus. Debye, from an
entirely different viewpoint, reaches the conclusion that the force
of attraction between molecules is inversely pro- portional to the
eighth power of the distance between them. From con- siderations of
this kind, it can be shown that the electric force near the surface
of an atom must decrease from a maximum value at the surface to
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
610 CHEMICAL REAClIONS ON SURFACES
a value one-half as great within a distance of about 0.3 x I O -
~ cm. In fact, an analysis of Debyes and other data shows that this
force decreases in about the same way on receding from the surface
of the atom, for molecules of widely differing type. In other
words, this distalice of 0-3 x 10~ is a nearly universal constant,
and in this way we get ;i much better conception of actual
conditions close to the surface of an atom than by assuming that
the force decreases with a power of the distance.
If magnetic forces exist within the atom it can be readily
calculated that these forces must decrease even more rapidly as the
distance from the atom increases.
I t must be said, therefore, that our present conception of the
structure of atoms and molecules makes it impossible for us to
conceive of any appreciable force which one atom or molecule can
exert directly on others at distances greater than two or three
Angstrom units ( I O - ~ cm.). Where effects are transmitted to
greater distances than these, it must be the result of a
transmission through and by atoms or molecules of matter. In view
of the structure of atoms from positive and negative particles, it
is clear that atoms should have the properties of a dielectric.
Thus, if we have a chain of atoms linked together by duplets-as,
for example, in the hydro- carbon chain of an organic compound-and
we bring a positively charged body near one end of the chain, the
electrons will be attracted and the nuclei repelled, so that a
certain displacement of these particles with respect to one another
will result. This effect is then transmitted with gradually
decreasing intensity from atom to atom throughout the length of the
chain, resulting in an accumulation of positive charge at the
opposite end of the chain. The chemical evidence indicates clearly
that effects of this kind are sometimes transmitted relatively
great distances. The many facts which have led some chemists to
assume polar valences, such as directed valences in organic
compounds, receive a simple explanation on the basis of these
transmitted effects.
In cases where atoms are not joined firmly to one another by the
sharing of duplets, we should never expect the transmission of
electric force to extend through more than about one atom. On this
basis, we are led to deny the existence of thick, stable, adsorbed
films of gas molecules such as those which were assumed in the
Nernst and in the Bodenstein-Fink theory of heterogeneous
reactions. If, for example, a surface is covered with a layer of
oxygen moZecuZes, then there should be little if any more tendency
for other molecules to form a second layer than there would be for
these molecules to remain in the surface of liquid oxyger, at the
same temperature. Thus, only when we have nearly saturated vapours
should we ever obtain films of gas molecules which exceed
monomolecular thickness.
The general opinion among colloid chemists and others who have
worked with adsorption effects, at least up to a few years ago,
seems to ha;e been that adsorbed films were usually of a thickness
of IOO to 1000 A. According to the views we have reached here, such
thick films cannot be regarded as the result of true adsorption,
but can result only from conden- sation in capillary spaces in
presence of nearly saturated vapours or are due to sorption or
solution. For example, it can be shown that glass, just like glue,
can sorb large quantities of water vapour, but this is a real
penetration of the water molecules into the solid material and is
not a strictly surface action.
There is no good reason for believing that it is only at low
pressures and high temperatures that adsorbed films are of
monomolecular thickness. The effect of catalytic poisons (as
studied, for example, by Faraday), sur-
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL REACTIONS ON SURFACES 61 I
face tension effects, the lubricating properties of thin oil
films, passivity phenomena in electrochemical actions, electrolytic
overvoltage, etc., all point unmistakably to the existence at
atmospheric pressure of stable films quite analogous to those
observed in high vacuum and at high tem- peratures.
Lord Rayleigh, in 1899, on the basis of some beautiful
experiments on surface tension, showed that the film of o$ve oil on
water contaminated with this substance has a thickness of 10 A.,
and is therefore probably of monomolecular thickness. This work was
later extended by Devaux, Labrouste, and others. These results were
of particular interest to the writer, because of their important
bearing on the question of the range of atomic and molecular forces
and the structure of adsorbed films in general.
Experimental results on the spreading of oils on water surfaces
have completely confirmed the views outlined above. The only oils
which spread on water are those whose molecules have active groups,
such as the -COOH, -OH, etc., which normally increase the
solubility of a sub- stance in water. The spreading therefore
occurs because the active group has an affinity for water, while
the hydrocarbon chain tends to remain in contact with other chains
of the same kind. The molecules on the surface must, therefore, be
oriented, so that the actual surface consists of the hydro- carbon
part of the molecules, while the active groups are all turned down-
ward towards the surface of the water. I t is evident that if we
have a series of substances having the same active group, but
different lengths of hydrocarbon chain, the number of molecules per
unit area in the oil film should remain about constant, while the
length of the molecule in the vertical direction, and therefore the
thickness of the film, should increase in proportion to the length
of the hydrocarbon chain. Numerous experi- ments have completely
verified these theoretical deducti0ns.l I n this way it becomes
possible to measure the lengths and cross-sections of the mole-
cules of oil films on surfaces, and to prove conclusively that the
films are not only monomolecular, but that orientation of the
molecules is a factor of vital importance in their formation. Very
accurate measurements of the forces involved in the formation of
these films and valuable additional information in regard to their
structural changes have recently been ob- tained in England by N.
K. Adam.2
Evidence that surface films are monomolecular and that the
molecules are oriented is also obtained from surface tension data
on pure liquids. While the spreading of oil films on water depends
upon the most active group in the molecule, the surface tension of
a liquid-which is a measure of the potential energy of its
surface-depends primarily on the least active group in the
molecule, for the group with the lowest stray field of force will
tend to form the actual surface layer, in order to make the
potential energy a minimum. An analysis of practically all
available published data on surface tension leads to a verification
oi this hypothesis.
The interfacial surface tension between two liquids, such as
water and mercury, or water and oil, gives, as W. B. Hardy has
shown, a measure of the energy changes involved in the formation of
the interface. W. D. Harkins has made numerous measurements of
interfacial surface tensions which show that the work done in the
formation of such interfaces is a
1 Langmuir, Met. Chem. Erkg., 15, 468 (1916) ; your . Amer. C h
e w sot., 39, 1848 (1917). A brief summary was published in Trans.
Faraday Society, 15, I (1920).
Yroc. Roy. SOC. A., 99, 336 (1921). JOUY. Amer. Chenz. Soc., 39,
354, 541 (1917).
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
612 CHEMICAL REACTIONS ON SURFACES
measure of the activity of the most active part of the molecule,
for the molecules become oriented at the interface,
A fourth method of determining the thickness of surface films
and proving that they are oriented in the surface, depends upon the
use of Gibb's thermodynamic equation, giving the total amount of
material ad- sorbed in the surface of a solution in terms of the
change in the surface tension of the solution as the concentration
of the solute is altered. By measuring the surface tension of
solutions at various concentrations it is thus possible to
determine the amount of material adsorbed per unit area, As the
concentration is increased, the amount adsorbed increases and
approaches a definite limit. The results show that in all such
cases the maximum amount adsorbed corresponds to that in a
monomolecular film.. I t is thus possible to determine the number
of molecules adsorbed per unit area and thus find the cross-section
of the molecules. The length is then obtained from the known volume
of the film. This method is appli- cable to adsorbed films on
liquids formed either from substances dissolved in the liquids, or
from substances present as vapour above the liquid. When a solution
has a lower surface tension than the pure solvent, the surface has
a monomolecular film of the dissolved substance, but where the
solution has a higher surface tension than the solvent the surface
of: the solution consists of a monomolecular film containing
nothing but pure solvent.
Direct experiments have also been made by the writer to
determine the maximum amount of gases that can be adsorbed by plane
surfaces of glass, mica, and p1atinum.l At ordinary temperatures,
with pressures of nitrogen, hydrogen, argon, carbon dioxide, etc.,
up to a few hundred bars at least, there is no measurable
adsorption by glass or mica-that is, less than I per cent. of the
surface is covered by a single layer of molecules. At the
temperature of liquid air, however, and at pressures of the order
of a. hundred bars, the surfaces become saturated by an adsorbed
film which never exceeds one molecule in thickness. The evidence is
that these films consist of modecudes and that primary valences are
not involved in their formation. The forces involved are
unquestionably the result of the stray field of force around the
molecule, and involve no radical rearrangement of the electrons.
The forces are probably very much like those involved in the
formation of substances containing water of crystallisation or
ammonia of crystallisation.
With a clean platinum surface which had been made catalytically
active, by bringing it into contact with a mixture of oxygen and
hydrogen at low pressures at a temperature of about 300' C., the
adsorption phenomena were totally different from those observed
with glass and mica, at least in the case of the gases hydrogen,
oxygen, and carbon monoxide. When small amounts of oxygen were
allowed to come in contact with the platinum surface, the oxygen
disappeared almost instantly, until the total amount adsorbed
corresponded to a monomolecular or monatomic film, and then no
further amount of this gas could be adsorbed, even with a great
increase in pressure. No trace of the oxygen could be pumped off by
heating the platinum in the best vacuum to 360".
If the platinum in this condition was allowed to come in contact
with hydrogen or carbon monoxide at low pressure, the oxygen film
was removed and water vapour or carbon dioxide was produced, even
at room tempera- ture, and then an additional amount of hydrogen or
carbon monoxide was
13!ou~ . Amer. Chem. SOC., 40, 1361 (1918).
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL REACTIONS ON SURFACES 613
adsorbed sufficient to produce a nionomolecular film of these
substances. The carbon monoxide film could be very gradually pumped
off at a temperature above 300'. No measurable amounts of nitrogen
or carbon dioxide were adsorbed by the platinum at any time.
These remarkably stable films on the platinum surface are of the
same type as the oxygen films adsorbed on tungsten surfaces.
Primary valences are unquestionably involved in their formation. I
n the case of the carbon monoxide, the carbon atom must be directly
attached to the platinum, while the oxygen is thus above the
carbon. The carbon monoxide mole- cules-if we can so speak of
them-are thus oriented on the surface, very much as the molecules
in an oil film. The experiments with platinum give a direct proof
that these stable films are of monomolecular thickness.
The evidence for the existence of monomolecular films is thus by
no means confined to experiments at low pressures for equally
striking evi- dence is furnished by the surface tension phenomena.
'The orientation of molecules in surface layers follows as a
necessary result from the conclusion that the range of atomic and
molecular forces is of the order of I A. The orientation in surface
films is a phenomenon with which we must constantly reckon, just as
we must consider structural relationships in the molecules of
organic compounds. Of course there are cases where the adsorbed
film consists of single atoms, or of various symmetrical molecules,
such as CH, or CCl,, where we do not need to consider orientation.
But wherever different parts of the surface of a molecule may be
assumed to have different pro- perties, we must take into account
the probability of orientation in all ad- sorption phenomena and
therefore in all catalytic actions on surfaces.
Mechrtnism uf Adsurption.--We have discussed the structure of
adsorbed films and the forces involved. Let us now consider the
mechanism by which these films form on a surface or disappear from
the surface.
When the adsorbed film of carbon monoxide on platinum gradually
dis- appears, on heating the metal to 300 in the highest vacuum, it
is logical to look upon this as an evaporation process. When a
filament of platinum, or tungsten, or other metal is heated to a
sufficiently high temperature in vacuum the material evaporates. If
the metal is placed in a uniformly heated enclosure, the
evaporation from the surface-which we may consider continues
unchanged-will be gradually ofTset by the return of atoms of metal
from the vapour which accumulates in the space. Finally, an equi-
librium is reached in which the rate of condensation of the vapour
is equal to the rate of evaporation.
If we can assume that all the atoms of the vapour which strike
the surface of the metal condense on the first collision, we may
calculate the rate of condensation from the vapour pressure by
means of the kinetic theory of gases. The formula usually given for
the rate of effusion of gases through small openings can readily be
put in the form
where M is the molecular weight of the vapour, R is the gas
constant,$ is the pressure of the vapour, and m is the rate at
which the gas molecules strike against the surface, in grams of
vapour per square centimetre per second. Expressing $J in bars, and
placing R = 83-15 x 106 ergs per degree, this reduces to
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
614 CHEMICAL REACTIONS ON SURFACES
This equation gives the rate at which the molecules of a vapour
strike against the surface. If every molecule condenses, and if we
have equi- librium, then the rate of evaporation must also be given
by this equation, so that we obtain a direct relation between the
vapour pressure of a sub- stance and its rate of evaporation in
perfect vacuum.
Experiments with many different metals have shown close
agreement between the vapour pressures determined in this way from
the rate of evaporation and the vapour pressures measured by
processes which involve the formation of saturated vapours. Knudsen
has made careful experi- ments of this kind with mercury vapour,
while A. S. Egerton 1 has carried out work with cadmium and zinc.
Their results indicate that every atom of vapour condenses.
Knudsen and R. W. Wood independently arrived at the ccnclusion
that mercury or cadmium atoms condense on a glass surface only if
this surface is cooled below a certain critical temperature. Below
this temperature, practically every atom is supposed to condense,
while at temperatures materially above this critical point not one
atom, out of thousands which strike the surface, condenses. This
conclusion is not only inherently im- probable in many ways, but is
not capable of accounting for numerous ex- perimental facts.
Woods*and Knudsens experiments are better explained by assuming
that all the atoms of cadmium and mercury which strike a glass
surface even at high temperature, condense on the surface, but that
at temperatures above the critical temperature, the atoms
re-evaporate before they have a chance to be struck by other atoms
of the vapour. The writer has discussed this question in detail in
a paper in the Physical Re- view, 8, 149 (1916), and subsequently
carried on experiments with cadmium vapour which demonstrate
conclusively that cadmium atoms evaporate rapidly from a clean
glass surface at room temperature. There is no real reason for
believing that cadmium may not also evaporate from glass at
temperatures only slightly above the critical temperature of - 90
C. cited by Wood. Since molten cadmium does not wet glass it is
clear that the forces between a cadmium atom and a glass surface
are much less than be- tween cadmium atoms, and the rate of
evaporation should therefore be much higher than from a cadmium
surface.
I n most cases of adsorption we are dealing with a solid surface
having a strong field of force, or a high potential energy per unit
area, while on this solid is condensed a substance whose molecules
possess a rather weak stray field of force. These are the
conditions when ordinary gases condense on cooled surfaces of glass
or metals. The forces which might tend to hold a second layer of
molecules are so weak that evaporation from the second layer occurs
at a rate high compared with that from the first. Only with nearly
saturated vapours, then, can a second layer form.
With cadmium and mercury vapours condensing on glass, however,
we have a case in which the evaporation from the second layer takes
place much more slowly than from the first layer. We see,
therefore, that a kind of instability necessarily results. There is
considerable difficulty in getting the first layer to form, because
the atoms tend to evaporate before the others are able to condense
on top of them or beside them. I f the first layer ever does form,
then the evaporation practically ceases and successive layers are
then formed with ease. This view seems to give a clear picture of
the mechanism of the formation of nuclei on which condensation
occurs. The formation of frost crystals on a greasy window pane, or
Mosers breath figures on glass, are illustrations of effects of
this kind.
Phil. Maz., 33, 33 (rgr7). PYOC. Nat. Acud. Sci., 3, 141
(1917).
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL RELICTIONS ON SURFACES
We might suppose that when molecules of a gas strike a surface,
only a certain fraction a condense on the surface, while the others
are reflected. Experimentally, however, it seems hard to find
examples where a is appreci- ably different from unity. Soddy,
Knudsen, and others have found, how- ever, that in heat conduction
from a solid surface to a gas at low pressure, the gas molecules
which strike the surface do not always reach thermal equilibrium
with the surface before leaving it. Knudsen has given the name
accommodation coefficient to the fraction which expresses the ratio
of the actual heat conduction to that calculated on the assumption
of heat equilibrium. I t is to be noted, however, that these
coefficients are usually of the order of magnitude of 0.8, and they
are determined under conditions in which the rate of evaporation of
the gas from the surface is unusually high. IVe should probably
therefore look upon these as rather exceptional cases, and, in
normal cases, unless we have definite evidence to the con- trary,
should assume that the coefficient a is unity.
When gas molecules of any kind strike a surface, we should
therefore not expect them to rebound elastically, but rather expect
them to condense. Adsorption is thus the result of the time lag
between condensation and evaporation. I n some cases the rate of
evaporation is so low that evapora- tion practically never occurs.
This is what happens, for example, when a catalyst is poisoned by
sulphur or arsenic compounds. I n other cases, the rate of
evaporation may be so high that the time that elapses between
condensation and evaporation may be of the order of 10 -I2 seconds
or even less.
2% Mechznism qf ChentzcnZ Xeactiom UIZ SzLrfuces.-The clean
surface of a solid crystalline body must consist of atoms or
molecules arranged in a surface lattice, or kind of checkerboard.
Non-crystalline bodies, such as glass, must have surfaces in which
the atoms are probably not in regular lattices. We may also have
surfaces which are porous, or consist of irregular filamentary
projections and interlocking chains of atoms or molecules. In such
cases the extent of the surface cannot be defined, except in a
purely arbitrary manner. Most finely divided catalysts, such as
platinum black, or activated charcoal, etc., must have structures
of great complexity, and it is probable that the atoms are attached
to each other in the form of branching chains so that there are
hardly any groups of as little as three or four atoms which are as
closely packed as they would be in the crystalline solid. I n order
to simplify our theoretical consideration of reactions on surfaces,
let us confine our attention mainly to reactions on plane surfaces.
If the principles in this case are well understood, it should then
be possible to extend the theory to the case of porous bodies.
I n general, we should look upon the surface of a catalyst as
consisting of a checkerboard in which some of the spaces are
vacant, while others are filled with atoms 01 molecules. Some of
these molecules, or atoms, may be so firmly attached that they do
not evaporate zlt an appreciable rate. Others leave the surface
from time to time, and the vacant spaces thus left are sooner or
later filIed by other molecules which strike the surface and
condense.
Xf we have a surface such as that of platinum, and we allow to
come in contact with it a gas, which forms an adsorbed film that
evaporates slowly, or not a t all, the surface is no longer a
platinum surface as far as possible interaction with other gas
molecules is concerned. The catalytic activity of the platinum has
thus been lost, or the catalyst has been poisoned. Arsenic,
sulphur, or phosphorus compounds have this effect, for the atoms of
these elements presumably combine directly with the atoms of
the
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
616 CHEMICAL REACTIONS ON SURFACES
platinum and do not evaporate at an appreciable rate. Cyanogen
and carbon monoxide have a similar, but more transient effect on
platinum, for only as long as these gases remain present in the gas
phase does the poisoning influence persist.
Faraday studied the effect of various substances in poisoning
the catalytic activity of platinum on the reaction between oxygen
and hydrogen. By boiling platinum foil in concentrated sulphuric
acid, and then washing with distilled water, it is brought into a
condition where it causes the combination of oxygen and hydrogen at
room temperature. The presence of carbon dioxide did not retard
this action, but a trace of carbon monoxide stopped the action
entirely, although on placing the platinum in a mixture of fresh
gas, free from monoxide, the reaction proceeded in a normal manner.
Hydrogen sulphide, or arsine, not only prevented the action while
they were present, but produced a permanent alteration in the
platinum, so that it was necessary to boil it again in acid before
it could be made active. The poisoning effect of oxygen on the
catalytic activity of tungsten at high temperature is of the
transient kind produced by carbon monoxide on platinum.
In the presence of a gas which has a poisoning effect on a
catalyst, the reaction velocity depends on that fraction of the
surface which is not covered by molecules of this gas. If the
temperature is high enough and the catalyst poison is of the kind
that has a transient effect, the adsorbed molecules evaporate at a
certain rate. If much of the gas is present, the vacant spaces thus
produced tend to be refilled by these molecules. The fraction of
the surface which is in an active condition is thus directly pro-
portional to the rate of evaporation of the film, and inversely
proportional to the partial pressure of the gas producing the
poisoning effect. We are thus led to an understanding of the
mechanism of the type of reaction which was explained by Bodenstein
and Fink by assuming adsorbed films having a thickness varying in
proportion to the pressure of a gas.
When gas molecules condense on a solid surface in such a way
that they are held on the surface by primaiy valence forces,
involving a rearrange- ment of their electrons, their chemical
properties become completely modified. I t is not surprising,
therefore, that in some cases such adsorbed films should be
extremely reactive, while in other cases they may be very inert to
outside influences. Thus oxygen adsorbed on platinum reacts readily
with hydrogen or carbon monoxide, while oxygen on tungsten, or
carbon monoxide on platinum, show very little tendency to react
with gases brought into contact with their surfaces. The specific
nature of the behaviour of these various films is quite consistent
with the theory that the adsorption depends on typical chemical
action. In many cases, especially where we deal with adsorption of
large molecules, the orientation of the molecules on the surface is
a factor of vital importance in determining the activity of the
surface towards reacting gases.
The reaction which takes place at the surface of a catalyst may
occur by interaction between molecules or atoms adsorbed in
adjacent spaces on the surface, or it may occur between an adsorbed
film and the atoms of the underlying solid, or again, it may take
place directly as a result of a collision between a gas molecule
and an adsorbed molecule or atom on the surface. This third kind of
action is perhaps indistinguishable from one in which the incident
gas molecules condense on top of those already on the surface, and
then react before they have a chance to evaporate.
When a surface is covered by different kinds of adsorbed
molecules distributed at random over the surface, we may expect in
general that
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL REACTIONS ON SURFACES
adsorbed molecules in adjacent spaces should be able to react
with one another at a rate which is proportional to the chance that
the given mole- cules shall lie in adjacent spaces. This kind of
mechanism has been dis- cussed at length by the writer in
connection with a study of the dissociation of hydrogen in contact
with a tungsten fi1ament.l When the hydrogen molecule strikes a
tungsten surface at high temperature, at least 68 per cent. of the
molecules condense on the surface and are held there as individual
atoms. After the action has proceeded for a time, the distribu-
tion of atomic hydrogen over the surface is given by the
probability laws. If O1 is the fraction of the surface covered by
this atomic hydro- gen, then the chance that any given elementary
space on the surface shall contain a hydrogen atom is O1. The
hydrogen atoms have a very strong field of force, since they are
unsaturated chemically (for the electrons are not arranged in
duplets). These atoms, therefore, have a relatively low rate of
evaporation from the surface. Two atoms in adjacent spaces on the
surface, however, may react with one another to form a hydrogen
molecule. This is chemically saturated, and has therefore a weak
field of force, so that it evaporates rapidly from the surface. The
rate of evapora- tion of molecular hydrogen is thus proportional to
its rate of formation from the atomic hydrogen, and this, in turn,
is proportional to 012, for the chance that two atoms shall lie in
adjacent spaces is proportional to the square of the chance than an
atom shall be in any given space. This statement of the problem
lends itself readily to mathematical treatment, and the equa- tions
that were obtained for the dependence of the reaction velocity on
the temperature and pressure are in full accord with experimental
facts over a temperature range from 1500' K. to 3500' K. and
pressures from 10 bars up to atmospheric pressure.
I t is probable that the decomposition of ammonia, and also the
forma- tion of ammonia in contact with solid catalysts, depends
upon similar interaction between adjacent adsorbed atoms. Reactions
of this sort should be extremely sensitive to the actual distances
between, and the arrangement of, the atoms in the surface of the
catalyst. If these atoms are a little too far apart, or if their
electrons are not sufficiently mobile to per- mit of the electron
rearrangement involved in surface reactions, the re- action will be
much retarded. I t is the opinion of the writer that these
differences in the geometrical arrangement of the atoms in the
surface is responsible for the " activation " of catalysts which is
brought about by the action that takes place upon them. For
example, if a plane surface of platinum be heated for the first
time in a mixture of hydrogen and oxygen, the temperature has to be
raised quite high before the reaction begins. When the reaction has
occurred, however, even in gases at very low pressures so that no
appreciable heating effect takes place, the catalyst becomes
modified and the reaction then proceeds, even at room temperature.
I n many cases, such effects are due to catalytic poisons, but
there is good evidence that the effect is frequently caused by
changes in the structure of the surface itself, brought about by
the reaction. This is particularly noticeable in the catalytic
oxidation of ammonia in contact with platinum wire. After the wire
has been used, the surface becomes very rough, and gradually a
disintegration of the wire occurs, because of the surface changes
taking place. The catalytic activity of the wire is very low when
first used, but becomes much greater after it has become activated
by the reaction itself.
your. Amev. Chena. SOC., 38, 1145 (1916).
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
618 CHEMICAL REACTIONS ON SURFACES
The changes that occur in the surface of the platinum under
these conditions seem to be exactly similar to those that are
caused by rapid fluctuation of temperature. When tantalum
filaments, or certain improperly made tungsten filaments, are run
in lamps on alternating current, the wire shows a tendency to
offset, but .this effect is entirely absent if the wire is heated
to the same temperature by continuous current. This offsetting
consists of a slipping of the crystals of the metal along the
boundary planes. I n extreme cases it leads to a nearly complete
disintegration of the structure of the metal. Experiments show that
this effect is directly dependent upon the rapidity of temperature
fluctuation. Anything that increases the rapidity of temperature
fluctuation, such as the introduction of hydrogen into the bulb,
increases the rate at which offsetting occurs, so that it is
possible in a few minutes to produce as much offsetting as would
other- wise occur during hundreds of hours. Under these extreme
conditions, the rate of cooling of the filament is of the order of
a million degrees per second.
If a Coolidge X-ray tube is operated exclusively with direct
current, even during manufacture, the surface of the target retains
its high polish, even after long use. A few minutes running with
alternating current roughens the surface of the target, and it is
iv$ ltnown that the focal spot in the target assumes an appearance
which is quite analogous to that of the platinum surface used as a
catalyst for the oxidation of ammonia. It is highly probable that
the cause is the same in both cases, namely, sudden fluctuations in
temperature between adjacent atoms in the: material.
In a surface of crystalline platinum, where the atoms are
presumably arranged in a definite surface lattice, the distances
between adsorbed atoms which occupy adjacent spaces is probably a
nearly fixed quantity, and in general it is unlikely that this
fortuitous spacing is the best adapted to the interaction between
the adsorbed molecules. When the surface atoms have been pushed
around and made to assume new positions arranged more or less at
random, the distances between adjacent adsorbed mole- cules vary
ova- a wide range, and some of these distances will be exactly
right for the reaction to occur at the highest possible speed. The
surface thus becomes composite, and there is then a relatively
small fraction of the surface at which the reaction occurs with
extreme rapidity, while over the larger part of the surface it
takes place at a very slow rate.
When a surface has become so roughened that it is porous, the
effective surface area increases, and the number of favourable
locations for the re- action to occur may become much greater.
There is good evidence, however, that the activation is not
merely due to an increase in the surface, for a surface which
becomes activated for one reaction may not become activated for
another reaction. For example, a plane surface of platinum, by a
single treatment in a hydrogen-oxygen mixture at low pressure, can
have its activity so much increased that the temperature at which
the reaction begins is lowered from 150 C. to room temperature, but
this increase in activity for the hydrogen-oxygen reaction is not
accompanied by any change in the velocity observed in the reaction
between carbon monoxide and oxygen.
The experiments seem to indicate that the reaction between
oxygen and hydrogen on platinum results from interaction between
adjacent adsorbed atoms, while the reaction between carbon monoxide
and oxygen takes place between oxygen atoms adsorbed on the surface
and carbon monoxide molecules from the gas phase which strike them.
This difference in mechanism probably accounts for the different
sensitiveness to surface conditions. It would also suggest that the
energy imparted to the in-
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
CHEMICAL REACTIONS ON SURFACES 619
dividual platinum atoms as a result of the reaction may be much
less in the carbon monoxide reaction than in the hydrogen reaction.
If this is so, the monoxide reaction should produce little change
in the surface, and thus should not activate the catalyst for the
hydrogen reaction.
The experimental evidence with carbon monoxide and oxygen on
platinum (described in detail in another paper presented at this
meeting), proves that nearly, but not quite all of the reaction
between these gases occurs during collision of carbon monoxide
molecules with the oxygen covered surface. In reactions of this
kind, which occur as the result of collisions, we may expect that
in some cases the exposure of the flanks I of an adsorbed film to
attack by colliding molecules may render them much more susceptible
to chemical action. For example, it is conceivable- although in
this particular case there is no experimental evidence for it-
that, if the whole surface of platinum were covered by oxygen
atoms, incident carbon monoxide molecules should be unable to
react, while if only a certain limited portion of the surface were
covered with oxygen, the mon- oxide molecules striking the oxygen
atoms close to the place where they are attached to the platinum,
might be able to react. In this case the oxygen film would be
removed progressively from its bounding edge inward. I t seems
quite possible that this kind of action may be involved in some of
the passivity phenomena observed with iron in electrochemical
action, and may also be effective in causing the szidden beginning
of the dissociation of hydrogen by a tungsten filament after small
traces of oxygen have been consumed by the filament.
If we consider catalytic surface reactions with more or less
complicated organic molecules, we should naturally expect that the
orientation of the molecules and steric hindrance effects should
become more important as factors in the mechanism of the reaction.
For example, when ethyl acetate is heated with different solid
catalysts, it may give-
A. CH3C02H + C,H4 B. CH3CHZCH3 + C02 C. CH3COCH3 + CO, + C,H,OH
+ C,H,. A. CH3C02H + C,H4 B. CH3CHZCH3 + C02 C. CH3COCH3 + CO, +
C,H,OH + C,H,.
In all these cases, the -COO- group is unquestionably directly
at- tached to the surface, while the rest of the hydrocarbon chain
is located above this group. I t is probable that the -COO- group
is attached to the surface by primary valences, so that the bonds
between these atoms disappear when the substance is adsorbed.
Depending upon the different manners in which interaction between
atoms and evaporation may occur, the resulting products differ.
Reaction A involves only a shift in the position of a hydrogen
nucleus, to allow the products to evaporate separ- ately. In
reaction B it is only necessary for a few electrons to shift their
positions. Reaction C involves interaction between two molecules
which must be adsorbed in adjacent positions in definite
geometrical relations to one another.
Reactions a t Boundaries of Phases.-Faraday observed that a
perfect crystal of sodium carbonate or sodium sulphate refuses to
effloresce until the surface is scratched or broken, and that the
efflorescence then spreads from the injured place. Similar
phenomena have been observed with copper sul- phate and other
crystals. In all such cases it is necessary to assume that the
reaction (dehydration) fakes jZace onZy a t the boundary between
two phases. Careful analysis2 shows that wherever we have to deal,
according
Experimental Researches, Everymans Library Edition, p. 109.
Langmuir, JOUY. Amer. Chem. SOC., 38, 2263-2267 (1916).
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
-
620 CHEMICAL REACTIONS ON SURFACES
to the phase rule, with separate phases of constant composition,
the re- action occurs only at the boundaries of phases. Thus in the
dissociation of calcium carbonate by heat, the carbon dioxide is
produced only at the boundary between the calcium carbonate and the
calcium oxide phases.
Dr. H. S. Taylor recently described experiments on the
preparation of copper and copper oxide catalysts in which he found
that there was a long
period of induction in the reduction of heated cupric oxide by
hydrogen. All the phenomena that he observed in connection with
this reaction are in accord with the view that this is another case
in which the reaction occurs only at the junction between phases.
In conventional nomenclature we may say that we have here an
example of autocatalysis, the metallic copper accelerating the
reaction. I t seems much more profitable, however, to analyse the
phenomena in terms of the probable mechanism.
The oxygen atoms in copper oxide are thoroughly saturated
chemically, which means in this case that they have taken up two
electrons from the copper atoms and have thus completed their
octets, and leave the copper in the form of ions. There is thus no
reason for expecting a strong ten-. dency to react with hydrogen at
moderate temperatures. In view of Taylors experiments, we conclude
that the oxygen ions in cupric oxide are in fact very inert towards
molecular hydrogen. Let us assume that, owing to some local
imperfection in the space lattice of the atoms of the copper oxide,
an ion of copper has taken up electrons and has formed a neutral
atom. I t is to be expected that such an atom should behave towards
hydrogen molecules like metallic copper or other metallic
substances. We have already seen that hydrogen molecules are
adsorbed by platinum and by tungsten (and therefore probably by
other metals) in the form of atoms. This action is presumably
caused by the attraction of the free electrons ,of the metal upon
the hydrogen nuclei. The hydrogen adsorbed by the copper in atomic
condition can then react with the oxygen ions merely by the
shifting of the hydrogen nuclei from the copper atoms to the oxygen
ions, the electrons being transferred to another copper atom. Each
hydro- gen molecule thus supplies two electrons and is capable of
converting an adjacent copper ion into a neutral atom. By such a
mechanism it is clear that the reaction could proceed only at the
junction between the phases.
Research Laboratory, GeneraZ EZectric Company,
Schenectady, N: E:, U. S. A.
Rochester Meeting, Amdr. Clicnt. SOC., April, 1921.
Publ
ished
on
01 Ja
nuar
y 19
22. D
ownl
oade
d by
Was
hing
ton
Stat
e Uni
vers
ity L
ibra
ries o
n 25
/10/
2014
04:
23:2
7.
View Article Online
